Merge pull request #88 from paul-gauthier/azure

Added args to configure openai module to access Azure

aider/coders/base_coder.py

@@ -53,8 +53,6 @@ class Coder:
         main_model,
         edit_format,
         io,
-        openai_api_key,
-        openai_api_base="https://api.openai.com/v1",
         **kwargs,
     ):
         from . import (
@@ -65,9 +63,6 @@ class Coder:
             WholeFileFunctionCoder,
         )
 
-        openai.api_key = openai_api_key
-        openai.api_base = openai_api_base
-
         if not main_model:
             main_model = models.GPT35_16k
 
@@ -630,6 +625,12 @@ class Coder:
         if functions is not None:
             kwargs["functions"] = self.functions
 
+        # we are abusing the openai object to stash these values
+        if hasattr(openai, "api_deployment_id"):
+            kwargs["deployment_id"] = openai.api_deployment_id
+        if hasattr(openai, "api_engine"):
+            kwargs["engine"] = openai.api_engine
+
         # Generate SHA1 hash of kwargs and append it to chat_completion_call_hashes
         hash_object = hashlib.sha1(json.dumps(kwargs, sort_keys=True).encode())
         self.chat_completion_call_hashes.append(hash_object.hexdigest())

aider/main.py

@@ -4,6 +4,7 @@ from pathlib import Path
 
 import configargparse
 import git
+import openai
 
 from aider import __version__, models
 from aider.coders import Coder
@@ -75,8 +76,27 @@ def main(args=None, input=None, output=None):
     model_group.add_argument(
         "--openai-api-base",
         metavar="OPENAI_API_BASE",
-        default="https://api.openai.com/v1",
-        help="Specify the OpenAI API base endpoint (default: https://api.openai.com/v1)",
+        help="Specify the openai.api_base (default: https://api.openai.com/v1)",
+    )
+    model_group.add_argument(
+        "--openai-api-type",
+        metavar="OPENAI_API_TYPE",
+        help="Specify the openai.api_type",
+    )
+    model_group.add_argument(
+        "--openai-api-version",
+        metavar="OPENAI_API_VERSION",
+        help="Specify the openai.api_version",
+    )
+    model_group.add_argument(
+        "--openai-api-deployment-id",
+        metavar="OPENAI_API_DEPLOYMENT_ID",
+        help="Specify the deployment_id arg to be passed to openai.ChatCompletion.create()",
+    )
+    model_group.add_argument(
+        "--openai-api-engine",
+        metavar="OPENAI_API_ENGINE",
+        help="Specify the engine arg to be passed to openai.ChatCompletion.create()",
     )
     model_group.add_argument(
         "--edit-format",
@@ -334,12 +354,19 @@ def main(args=None, input=None, output=None):
 
     main_model = models.Model(args.model)
 
+    openai.api_key = args.openai_api_key
+    for attr in ("base", "type", "version", "deployment_id", "engine"):
+        arg_key = f"openai_api_{attr}"
+        val = getattr(args, arg_key)
+        if val is not None:
+            mod_key = f"api_{attr}"
+            setattr(openai, mod_key, val)
+            io.tool_output(f"Setting openai.{mod_key}={val}")
+
     coder = Coder.create(
         main_model,
         args.edit_format,
         io,
-        args.openai_api_key,
-        args.openai_api_base,
         ##
         fnames=args.files,
         pretty=args.pretty,

docs/faq.md

@@ -50,12 +50,48 @@ Adopting new LLMs will probably require a similar effort to tailor the
 prompting and edit formats.
 
 That said, aider does provide some features to experiment with other models.
+
+### Azure
+
+Aider can be configured to connect to the OpenAI models on Azure.
+Aider supports the configuration changes specified in the
+[official openai python library docs](https://github.com/openai/openai-python#microsoft-azure-endpoints).
+You should be able to run aider with the following arguments to connect to Azure:
+
+```
+$ aider \
+    --openai-api-type azure \
+    --openai-api-key your-key-goes-here \
+    --openai-api-base https://example-endpoint.openai.azure.com \
+    --openai-api-version 2023-05-15 \
+    --openai-api-deployment-id deployment-name \
+    ...
+```
+
+You could also store those values in an `.aider.conf.yml` file in your home directory:
+
+```
+openai-api-type: azure
+openai-api-key: your-key-goes-here
+openai-api-base: https://example-endpoint.openai.azure.com
+openai-api-version: 2023-05-15
+openai-api-deployment-id: deployment-name
+```
+
+See the
+[official Azure documentation on using OpenAI models](https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?tabs=command-line&pivots=programming-language-python)
+for more information on how to populate the above configuration values.
+
+### Other LLMs
+
 If you can make the model accessible via an OpenAI compatible API,
 you can use `--openai-api-base` to connect to a different API endpoint.
 
 Here are some
 [GitHub issues which may contain relevant information](https://github.com/paul-gauthier/aider/issues?q=is%3Aissue+%22openai-api-base%22+).
 
+### Local LLMs
+
 [LocalAI](https://github.com/go-skynet/LocalAI)
 and
 [SimpleAI](https://github.com/lhenault/simpleAI)

tests/test_coder.py

@@ -28,7 +28,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Mock the git repo
         mock = MagicMock()
@@ -62,9 +62,7 @@ class TestCoder(unittest.TestCase):
         files = [file1, file2]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, None, io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, None, io=InputOutput(), fnames=files)
 
         content = coder.get_files_content().splitlines()
         self.assertIn("file1.txt", content)
@@ -75,7 +73,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         mock = MagicMock()
         mock.return_value = set(["file1.txt", "file2.py"])
@@ -101,7 +99,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         mock = MagicMock()
         mock.return_value = set(["file1.txt", "other/file1.txt"])
@@ -117,7 +115,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Mock the send method to set partial_response_content and return False
         def mock_send(*args, **kwargs):
@@ -137,7 +135,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Mock the send method to set partial_response_content and return False
         def mock_send(*args, **kwargs):
@@ -157,7 +155,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Mock the send method to set partial_response_content and return False
         def mock_send(*args, **kwargs):
@@ -179,7 +177,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Set up the mock to raise RateLimitError on
         # the first call and return None on the second call
@@ -201,7 +199,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Set up the mock to raise ConnectionError on the first call
         # and return None on the second call
@@ -230,9 +228,7 @@ class TestCoder(unittest.TestCase):
         files = [file1, file2]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, None, io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, None, io=InputOutput(), fnames=files)
 
         def mock_send(*args, **kwargs):
             coder.partial_response_content = "ok"
@@ -258,9 +254,7 @@ class TestCoder(unittest.TestCase):
         files = [file1, file2]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, None, io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, None, io=InputOutput(), fnames=files)
 
         def mock_send(*args, **kwargs):
             coder.partial_response_content = "ok"
@@ -290,9 +284,7 @@ class TestCoder(unittest.TestCase):
         files = [file1]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, None, io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, None, io=InputOutput(), fnames=files)
 
         def mock_send(*args, **kwargs):
             coder.partial_response_content = "ok"
@@ -320,7 +312,6 @@ class TestCoder(unittest.TestCase):
             models.GPT4,
             None,
             io=InputOutput(encoding=encoding),
-            openai_api_key="fake_key",
             fnames=files,
         )
 
@@ -349,7 +340,7 @@ class TestCoder(unittest.TestCase):
         mock_io = MagicMock()
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(models.GPT4, None, mock_io, openai_api_key="fake_key")
+        coder = Coder.create(models.GPT4, None, mock_io)
 
         # Set up the mock to raise InvalidRequestError
         mock_chat_completion_create.side_effect = openai.error.InvalidRequestError(
@@ -393,7 +384,6 @@ class TestCoder(unittest.TestCase):
             models.GPT4,
             None,
             io=InputOutput(),
-            openai_api_key="fake_key",
             fnames=[str(tempdir / filenames[0])],
         )
 

tests/test_commands.py

@@ -29,7 +29,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         # Call the cmd_add method with 'foo.txt' and 'bar.txt' as a single string
@@ -44,7 +44,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         # Create some test files
@@ -70,7 +70,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         # Call the cmd_add method with a non-existent file pattern
@@ -84,7 +84,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         # Create a directory and add files to it
@@ -117,7 +117,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         subdir = Path("subdir")
@@ -144,7 +144,7 @@ class TestCommands(TestCase):
         io = InputOutput(pretty=False, yes=True)
         from aider.coders import Coder
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         # Create a new file foo.bad which will fail to decode as utf-8
@@ -159,7 +159,7 @@ class TestCommands(TestCase):
         # Initialize the Commands and InputOutput objects
         io = InputOutput(pretty=False, yes=True)
 
-        coder = Coder.create(models.GPT35, None, io, openai_api_key="deadbeef")
+        coder = Coder.create(models.GPT35, None, io)
         commands = Commands(io, coder)
 
         commands.cmd_add("foo.txt bar.txt")

tests/test_editblock.py

@@ -248,9 +248,7 @@ These changes replace the `subprocess.run` patches with `subprocess.check_output
         files = [file1]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, "diff", io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, "diff", io=InputOutput(), fnames=files)
 
         def mock_send(*args, **kwargs):
             coder.partial_response_content = f"""
@@ -290,7 +288,6 @@ new
             models.GPT4,
             "diff",
             io=InputOutput(dry_run=True),
-            openai_api_key="fake_key",
             fnames=files,
             dry_run=True,
         )

tests/test_wholefile.py

@@ -292,9 +292,7 @@ after b
         files = [file1]
 
         # Initialize the Coder object with the mocked IO and mocked repo
-        coder = Coder.create(
-            models.GPT4, "whole", io=InputOutput(), openai_api_key="fake_key", fnames=files
-        )
+        coder = Coder.create(models.GPT4, "whole", io=InputOutput(), fnames=files)
 
         # no trailing newline so the response content below doesn't add ANOTHER newline
         new_content = "new\ntwo\nthree"